Stairstep effect on graphs

[dune]

I'm trying to get a nudge in the right direction in resolving a "stairstep" effect in some (not all) of my graphs. The effect varies depending what time range is used which indicates a problem with how values are averaged over time. Since the RRA settings are global and I cannot find differences in the graph templates, I'm not sure where else to look.

Here are some images illustrating the problem:

1-week period (OK)



2-week period (note stairstep on one graph but not the other)

[mcutting]

Check your RRA settings. Looks like an imported template could have changed these settings.

[gandalf]

There's an option in the graph template for smoothing, IMHO. Or does it come in 087c???
Reinhard

[dune]

Here's how my RRA's look now but I'm not sure if these have changed from default:

Code: Select all

Name                           Steps  Rows  Timespan**
Hourly (1 Minute Average)      1      500   14400
Daily (5 Minute Average)       1      600   86400
Weekly (30 Minute Average)     6      700   604800
Monthly (2 Hour Average)       24     775   2678400
Yearly (1 Day Average)         288    797   33053184

Also, wouldn't the RRA settings affect all graphs equally?

Regarding smoothing I'm not sure that's the source of the problem. The 2-week graph that doesn't have the effect seems to be pulling more discrete values.

Is it possible for an individual graph template to have different RRA settings than the globals?

[gandalf]

Please see both Graph Templates and check the "--slope-mode" setting
Reinhard

[dune]

Please see both Graph Templates and check the "--slope-mode" setting
Reinhard

Both templates have the Slope Mode option checked. In fact, aside from some naming and the color both templates appear to be identical. Could there be a template setting not accessible from the web interface?

[gandalf]

Please visit both graphs at Graph Management. Switch to debug and post both rrdtool graph statements
Reinhard

[dune]

Code: Select all

RRDTool Command:

c:/cacti/rrdtool.exe graph - \
--imgformat=PNG \
--start=-86400 \
--end=-60 \
--title="prd-edge-fw01 - CPU Utilization" \
--rigid \
--base=1000 \
--height=120 \
--width=500 \
--alt-autoscale-max \
--lower-limit=0 \
--vertical-label="% utilization" \
--slope-mode \
--font TITLE:8:c:/cacti/fonts/trebuc.ttf \
--font AXIS:8: \
--font LEGEND:8: \
--font UNIT:8: \
DEF:a="D\:/content/prod-cacti-v1/docs/rra/prd-edge-fw01_5sec_cpu_580.rrd":5sec_cpu:LAST \
DEF:b="D\:/content/prod-cacti-v1/docs/rra/prd-edge-fw01_5sec_cpu_580.rrd":5sec_cpu:AVERAGE \
AREA:a#D00000:"Utilization"  \
GPRINT:a:LAST:"Current\:%8.0lf"  \
GPRINT:b:AVERAGE:"Average\:%8.0lf"  \
GPRINT:b:MAX:"Maximum\:%8.0lf" 

RRDTool Says:

OK

Code: Select all

RRDTool Command:

c:/cacti/rrdtool.exe graph - \
--imgformat=PNG \
--start=-86400 \
--end=-60 \
--title="prd-edge-fw01 - Current Connections" \
--rigid \
--base=1000 \
--height=120 \
--width=500 \
--alt-autoscale-max \
--lower-limit=0 \
--vertical-label="connections" \
--slope-mode \
--font TITLE:8:c:/cacti/fonts/trebuc.ttf \
--font AXIS:8: \
--font LEGEND:8: \
--font UNIT:8: \
DEF:a="D\:/content/prod-cacti-v1/docs/rra/prd-edge-fw01_pixconnections_340.rrd":pixconnections:LAST \
DEF:b="D\:/content/prod-cacti-v1/docs/rra/prd-edge-fw01_pixconnections_340.rrd":pixconnections:AVERAGE \
AREA:a#0000C8:"Connections"  \
GPRINT:a:LAST:"Current\:%8.0lf"  \
GPRINT:b:AVERAGE:"Average\:%8.0lf"  \
GPRINT:b:MAX:"Maximum\:%8.0lf" 

RRDTool Says:

OK

[dune]

I did a CSV export on both graphs and notice that one is using hourly-average data and the other is using 5-hour averages. The OK graph has a step of 3600 while the other has one of 17280.

This explains the effect but not what is causing it.

[dune]

Any further suggestions? I've tried blowing away and re-creating the graphs and data sources but it is still occurring.

[TheWitness]

A step of 3600 is an hourly update. The other one is several hours. I think you need to size your RRD's appropriately. The rest may be your imagination.

TheWitness

[dune]

Ok, I ran an rrdtool info on the RRD files behind these and one step is 60 (bad graph) while the other is 300 (good graph).

The data source templates are both set to 60... should they be 300?

[TheWitness]

For a step of 60 to work, you would have to have a 1 minute poller interval. I prefer 300 (5 minutes) unless it's something that rolls over in 5 minutes like Gig interfaces forced to use snmpv1.

I use the Realtime Plugin if I need to dig in.

TheWitness

[dune]

My goal is a 1-minute polling interval... at least that's what I have in the Cacti settings. The poller task runs every 1 minute as well.

[TheWitness]

Well, you should audit your RRDfiles by doing something like the following:

Code: Select all

#!/bin/ksh
cd /var/www/html/cacti
find . -name \*.rrd -exec rrdtool info {} | grep -i step \;

TheWitness